Yes. It seems the procedure that we use to create the vray instances from meshes, was creating some memory instability in VRay 2.0, although it was working fine in VRay 1.5. We don't know if it would be a VRay issue or our fault (everything in the Forest code seems to be correct).
Unfortunately, the crash it is not easy to reproduce. We needed to run automatic tests for several hours, introducing small changes in the code on each cyle, to find what it was the problem.
Anyways, we have included a workaround in Forest 3.4.1 that must solve this issue. As soon as Vlado will be back on the Chaosgroup forum, we will send to him all the information that we have collected in our tests.